[Fetch API] Implement Request.formData and Response.formData.

third_party/WebKit/Source/modules/fetch/MultipartParser.cpp

+// Copyright 2016 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "modules/fetch/MultipartParser.h"
+
+#include "platform/HTTPNames.h"
+#include "public/platform/Platform.h"
+
+#include <algorithm>
+#include <utility>
+
+namespace blink {
+
+namespace {
+
+constexpr char kCloseDelimiterSuffix[] = "--\r\n";
+constexpr size_t kCloseDelimiterSuffixSize = WTF_ARRAY_LENGTH(kCloseDelimiterSuffix) - 1u;
+constexpr size_t kDashBoundaryOffset = 2u; // The length of "--".
+constexpr char kDelimiterSuffix[] = "\r\n";
+constexpr size_t kDelimiterSuffixSize = WTF_ARRAY_LENGTH(kDelimiterSuffix) - 1u;
+
+bool octetsNeedDecoding(const WebURLResponse& response)
+{
+    const AtomicString& encoding = response.toResourceResponse().httpHeaderFields().get(HTTPNames::Content_Transfer_Encoding);
+
+    // Decoding is not needed if a transfer encoding is not used.
+    if (encoding.isNull())
+        return false;
+
+    // Decoding is not needed if a no-op transfer encoding is used.
+    if (equalIgnoringCase(encoding, "binary")
+        || equalIgnoringCase(encoding, "7bit")
+        || equalIgnoringCase(encoding, "8bit")) {
+        return false;
+    }
+
+    return true;
+}
+
+} // namespace
+
+MultipartParser::Matcher::Matcher() = default;
+
+MultipartParser::Matcher::Matcher(const char* data, size_t size, size_t capacity)
+    : m_capacity(capacity), m_data(data), m_size(size) {}
+
+bool MultipartParser::Matcher::appendIfExpected(char value)
+{
+    DCHECK_LT(m_size, m_capacity);
+    if (value != m_data[m_size])
+        return false;
+    ++m_size;
+    return true;
+}
+
+bool MultipartParser::Matcher::appendIfExpected(const char* first, const char* last)
+{
+    while (first < last) {
+        if (!appendIfExpected(*first++))
+            return false;
+    }
+    return true;
+}
+
+void MultipartParser::Matcher::clear()
+{
+    m_size = 0u;
+}
+
+MultipartParser::MultipartParser(Vector<char> boundary, Client* client)
+    : m_client(client)
+    , m_delimiter(std::move(boundary))
+    , m_state(ParsingPreamble)
+{
+    // The delimiter consists of "\r\n" and a dash boundary which consists of
+    // "--" and a boundary.
+    m_delimiter.prepend("\r\n--", 4u);
+    m_bufferedBytes = delimiterMatcher(kDashBoundaryOffset);
+}
+
+bool MultipartParser::appendData(const char* bytes, size_t size)
+{
+    DCHECK_NE(Finished, m_state);
+    DCHECK_NE(Cancelled, m_state);
+
+    const char* bytesEnd = bytes + size;
+
+    while (bytes < bytesEnd) {
+        switch (m_state) {
+        case ParsingPreamble:
+            // Parse either a preamble and a delimiter or a dash boundary.
+            parseDelimiter(&bytes, bytesEnd);
+            if (!m_bufferedBytes.isComplete() && bytes < bytesEnd) {
+                // Parse a preamble (by ignoring it) and parse a delimiter.
+                m_bufferedBytes.clear();
+                parseDataAndDelimiter(&bytes, bytesEnd);
+            }
+            if (m_bufferedBytes.isComplete()) {
+                // Prepare for a delimiter suffix.
+                m_bufferedBytes = delimiterSuffixMatcher();
+                m_state = ParsingDelimiterSuffix;
+            }
+            break;
+
+        case ParsingDelimiterSuffix:
+            // Parse transport padding and "\r\n" after a delimiter.
+            // This can happen after either a preamble or part octets.
+            if (m_bufferedBytes.empty())
+                parseTransportPadding(&bytes, bytesEnd);
+            while (bytes < bytesEnd) {
+                if (!m_bufferedBytes.appendIfExpected(*bytes++))
+                    return false;
+                if (m_bufferedBytes.isComplete()) {
+                    // Prepare for part header fields.
+                    m_state = ParsingPartHeaderFields;
+                    break;
+                }
+            }
+            break;
+
+        case ParsingPartHeaderFields: {
+            // Parse part header fields (which ends with "\r\n") and an empty
+            // line (which also ends with "\r\n").
+            // This can happen after a delimiter and a delimiter suffix which
+            // can happen after either a preamble or part octets.
+            WebURLResponse response;
+            if (parseHeaderFields(&bytes, bytesEnd, &response)) {
+                // Decoding is not implemented.
+                if (octetsNeedDecoding(response))
+                    return false;
+                // Prepare for part octets.
+                m_bufferedBytes = delimiterMatcher();
+                m_state = ParsingPartOctets;
+                m_client->partHeaderFieldsInMultipartReceived(response.toResourceResponse());
+            }
+            break;
+        }
+
+        case ParsingPartOctets: {
+            // Parse part octets and a delimiter.
+            // This can happen only after part header fields.

[yhirano, 2016/09/21 09:02:57]

Can you tell me what "This can happen" mean?

[e_hakkinen, 2016/09/22 22:27:16]

It refers to this parsing part octets and a delimiter. I changed the wording.

+            bool hasInitiallyBufferedBytes = !m_bufferedBytes.empty();
+            const char* octetsBegin = bytes;
+            parseDelimiter(&bytes, bytesEnd);
+            if (!m_bufferedBytes.isComplete() && bytes < bytesEnd) {
+                if (!m_bufferedBytes.empty() && hasInitiallyBufferedBytes) {
+                    // The buffered bytes turned out to be octet bytes instead
+                    // of delimiter bytes. Additionally, some of the buffered
+                    // bytes are from the previous call and are therefore not
+                    // in the range [octetsBegin, bytesEnd[.
+                    m_client->partDataInMultipartReceived(m_bufferedBytes.data(), m_bufferedBytes.size());
+                    octetsBegin = bytes;
+                }
+                if (m_state == ParsingPartOctets) {

[yhirano, 2016/09/21 09:02:57]

[optional] Moving the if statement to just after the notification dispatching is
more intuitive.

[yhirano, 2016/09/21 09:02:57]

if (m_state != ParsingPartOctets)
  break;

would be easier to read.

[e_hakkinen, 2016/09/22 22:27:16]

Done.

+                    m_bufferedBytes.clear();
+                    parseDataAndDelimiter(&bytes, bytesEnd);
+                    const char* octetsEnd = bytes - m_bufferedBytes.size();
+                    if (octetsEnd > octetsBegin)

[yhirano, 2016/09/21 09:02:57]

You are using the |begin < end| pattern in this file except here, and being
consistent is generally good.

[e_hakkinen, 2016/09/22 22:27:16]

Done.

+                        m_client->partDataInMultipartReceived(octetsBegin, static_cast<size_t>(octetsEnd - octetsBegin));
+                }
+            }
+            if (m_state == ParsingPartOctets && m_bufferedBytes.isComplete()) {
+                m_state = ParsingDelimiterOrCloseDelimiterSuffix;
+                m_client->partDataInMultipartFullyReceived();
+            }
+            break;
+        }
+
+        case ParsingDelimiterOrCloseDelimiterSuffix:
+            // Determine whether this is a delimiter suffix or a close
+            // delimiter suffix.
+            // This can happen only after part octets.
+            if (*bytes == '-') {
+                // Prepare for a close delimiter suffix.
+                m_bufferedBytes = closeDelimiterSuffixMatcher();
+                m_state = ParsingCloseDelimiterSuffix;
+            } else {
+                // Prepare for a delimiter suffix.
+                m_bufferedBytes = delimiterSuffixMatcher();
+                m_state = ParsingDelimiterSuffix;
+            }
+            break;
+
+        case ParsingCloseDelimiterSuffix:
+            // Parse "--", transport padding and "\r\n" after a delimiter
+            // (a delimiter and "--" constitute a close delimiter).
+            // This can happen only after part octets.
+            for (;;) {
+                if (m_bufferedBytes.size() == 2u)
+                    parseTransportPadding(&bytes, bytesEnd);
+                if (bytes >= bytesEnd)
+                    break;
+                if (!m_bufferedBytes.appendIfExpected(*bytes++))
+                    return false;
+                if (m_bufferedBytes.isComplete()) {
+                    // Prepare for an epilogue.
+                    m_state = ParsingEpilogue;
+                    break;
+                }
+            }
+            break;
+
+        case ParsingEpilogue:
+            // Parse an epilogue (by ignoring it).
+            // This can happen only after a delimiter and a close delimiter
+            // suffix which can happen only after part octets.
+            return true;
+
+        case Cancelled:
+        case Finished:
+            // The client changed the state.
+            return true;
+        }
+    }
+
+    DCHECK_EQ(bytesEnd, bytes);
+
+    return true;
+}
+
+void MultipartParser::cancel()
+{
+    m_state = Cancelled;
+}
+
+bool MultipartParser::finish()
+{
+    DCHECK_NE(Cancelled, m_state);
+
+    State initialState = m_state;
+
+    if (m_state == ParsingPartOctets && !m_bufferedBytes.empty()) {
+        // The end of append bytes looked like a delimiter but was not a full
+        // one, after all. Treat the those bytes as octet bytes.
+        m_client->partDataInMultipartReceived(m_bufferedBytes.data(), m_bufferedBytes.size());
+    }
+    m_state = Finished;
+
+    switch (initialState) {
+    case ParsingCloseDelimiterSuffix:
+        // Require a full close delimiter consisting of a delimiter and "--"
+        // but ignore missing or partial "\r\n" after that.
+        return m_bufferedBytes.size() >= 2u;
+    case ParsingEpilogue:
+    case Finished:
+        return true;
+    default:
+        return false;
+    }
+}
+
+MultipartParser::Matcher MultipartParser::closeDelimiterSuffixMatcher() const
+{
+    return Matcher(kCloseDelimiterSuffix, 0u, kCloseDelimiterSuffixSize);
+}
+
+MultipartParser::Matcher MultipartParser::delimiterMatcher(size_t size) const
+{
+    return Matcher(m_delimiter.data(), size, m_delimiter.size());
+}
+
+MultipartParser::Matcher MultipartParser::delimiterSuffixMatcher() const
+{
+    return Matcher(kDelimiterSuffix, 0u, kDelimiterSuffixSize);
+}
+
+void MultipartParser::parseDataAndDelimiter(const char** bytesPointer, const char* bytesEnd)
+{
+    DCHECK(m_bufferedBytes.empty());
+
+    // Search for a complete delimiter within the bytes.
+    const char* delimiterBegin = std::search(*bytesPointer, bytesEnd, m_delimiter.begin(), m_delimiter.end());
+    if (delimiterBegin != bytesEnd) {
+        // A complete delimiter was found. The bytes before that are octet
+        // bytes.
+        const char* delimiterEnd = delimiterBegin + m_delimiter.size();
+        if (!m_bufferedBytes.appendIfExpected(delimiterBegin, delimiterEnd))

[yhirano, 2016/09/21 09:02:57]

bool matched = m_bufferedBytes.appendIfExpected(delimiterBegin, delimiterEnd);
DCHECK(matched);

[e_hakkinen, 2016/09/22 22:27:16]

Done.

+            NOTREACHED();
+        DCHECK(m_bufferedBytes.isComplete());
+        *bytesPointer = delimiterEnd;
+    } else {
+        // Search for a partial delimiter in the end of the bytes.
+        size_t size = static_cast<size_t>(bytesEnd - *bytesPointer);
+        delimiterBegin = bytesEnd - std::min(m_delimiter.size() - 1u, size);
+        for (;;) {

[yhirano, 2016/09/21 09:02:57]

How about

for (const char* begin = ...; begin < bytesEnd; ++begin) {
  ...
}

?

[e_hakkinen, 2016/09/22 22:27:16]

Done. That decreases the number of times the loop is repeated by one, but that
is OK.

+            if (m_bufferedBytes.appendIfExpected(delimiterBegin, bytesEnd))
+                break;
+            m_bufferedBytes.clear();
+            DCHECK_LT(delimiterBegin, bytesEnd);
+            ++delimiterBegin;
+        }
+        // A partial delimiter (consisting of zero or more bytes) was found in
+        // the end of bytes. The partial delimiter bytes are buffered for now.
+        // The bytes before that are definitely octet bytes.
+        *bytesPointer = bytesEnd;
+    }
+
+    DCHECK(m_bufferedBytes.isComplete() || *bytesPointer == bytesEnd);
+}
+
+void MultipartParser::parseDelimiter(const char** bytesPointer, const char* bytesEnd)
+{
+    DCHECK(!m_bufferedBytes.isComplete());
+    while (*bytesPointer < bytesEnd && m_bufferedBytes.appendIfExpected(*(*bytesPointer))) {
+        ++(*bytesPointer);
+        if (m_bufferedBytes.isComplete())
+            break;
+    }
+}
+
+bool MultipartParser::parseHeaderFields(const char** bytesPointer, const char* bytesEnd, WebURLResponse* response)
+{
+    // Combine the current bytes with buffered header bytes if needed.
+    const char* headerBytes = *bytesPointer;
+    size_t headerSize = static_cast<size_t>(bytesEnd - *bytesPointer);
+    if (!m_bufferedHeaderBytes.isEmpty()) {
+        m_bufferedHeaderBytes.append(headerBytes, headerSize);
+        headerBytes = m_bufferedHeaderBytes.data();
+        headerSize = m_bufferedHeaderBytes.size();
+    }
+
+    size_t end = 0u;
+    if (!Platform::current()->parseMultipartHeadersFromBody(headerBytes, headerSize, response, &end)) {
+        // Store the current header bytes for the next call unless that has
+        // already been done.
+        if (m_bufferedHeaderBytes.isEmpty())
+            m_bufferedHeaderBytes.append(headerBytes, headerSize);
+        *bytesPointer = bytesEnd;
+        return false;
+    }
+    m_bufferedHeaderBytes.clear();
+    *bytesPointer = bytesEnd - (headerSize - end);
+
+    return true;
+}
+
+void MultipartParser::parseTransportPadding(const char** bytesPointer, const char* bytesEnd) const
+{
+    while (*bytesPointer < bytesEnd && (*(*bytesPointer) == '\t' || *(*bytesPointer) == ' '))
+        ++(*bytesPointer);
+}
+
+DEFINE_TRACE(MultipartParser)
+{
+    visitor->trace(m_client);
+}
+
+} // namespace blink